A running database is no longer reported as down after a node failover.

Detail name	Value
Changelog Number	32942
Type	Bug
Status	Open
Affected Versions	Exasol 7.1.0

Description

After a node failure, dwad could keep a node in its active node list after that node was gone. Internal state updates then waited for that node until they timed out. A database restarting at that moment was never recorded as accepting client connections. It served clients normally while dwad and ConfD reported its connection state as down. Scheduled backups depend on that state and stopped without an error.

Workaround

Restart the database.

Fix

dwad now requests the authoritative node list from cored when an internal state update fails, and drops nodes that left the cluster. Updates no longer wait for a node already known to be unreachable.